[Rebuild] Getting underway before 1st April

[Simon Poole]

Am 01.02.2012 19:47, schrieb Frederik Ramm:
> .........
> I would also feel very relieved if somebody would start from a clean 
> slate and come up with an algorithm to determine "what is clean", run 
> it on a full history file and compare results with what I have. I 
> mean, this is just stuff I hacked up on a weekend! There surely should 
> be *some* quality assurance?
I've in general not seen any fundamental differences to WTFE in both the 
statistics that I calculate and the underlying data I use for cleanmap 
(which was essentially seeded more or less the same way as Frederik 
proposes for the transition).

What I currently don't do that you do is separate out any harmless 
edits, including moved tagless nodes (but that is more or less trivial).

I've not come across any differences in the data which didn't boil down 
to the above differences and I wouldn't have any issues at all using the 
logic from WTFE for the transition.

>
>> - and that the resulting map would look almost exactly the same as the
>> CleanMap Simon is publishing now?
>
> I think that Simon is making a number of assumptions that we would not 
> get away with, notably that if a way is clean then we can always use 
> all its nodes no matter what. So the map paints a rosier picture than 
> what we will have.
I naturally have the underlying data to do this, but actually using it 
in the map would imply changing the linestring building logic in 
osm2pgsql or rebuilding the affected linestrings after each update, both 
don't make very much sense for the remaining 60 days,

I further don't change tainted tags on a object to something clean, but 
visually that is only a minor effect.

Simon